Meta-heuristic Solution for Dynamic Association Control in Virtualized Multi-rate WLANs Dawood Sajjadi, Maryam Tanha, Jianping Pan Department of Computer Science, University of Victoria, BC, Canada November 9 th 2016, Dubai, UAE (LCN 2016)
Outline 1. Background 2. Motivations 3. Big Picture 4. Related Work 5. Problem Formulation 6. Meta-heuristic Solution 7. Evaluation Scenarios 8. Results 9. Summary 1/13
Background ~100 million people with Cell Phones and No Electricity 2/13
Motivation It happens many times! There are so many Access Points around us, but we have a Poor Wi-Fi Connection. 3/13
Big Picture ISP-1 ISP-2 ISP-3 Controller Virtualized/Centralized WLAN Infrastructure Building a common WLAN infrastructure for Cooperative ISPs (Centralized Management) AP2 SSID-3 AP3 SSID-2 Increasing the Service Coverage and Promoting QoS/QoE for the Customers SSID-2 Providing Proportional Fairness (PF) among the Clients AP1 AP4 Protocol-Independent Solution for Adaptive Association Control 4/13
Related Works Greedy co-channel APs Downlink traffic Compleity Running time No comparison with optimal solution/other schemes Cooperative Association Control Single Channel WLANs Disregarding the impact of interference Simultaneous association to multiple APs 1. W. Li, S. Wang, Y. Cui, X. Cheng, R. Xin, M. A. Al-Rodhaan, and A. Al-Dhelaan, AP association for proportional fairness in multi-rate WLANs IEEE/ACM Trans. on Networking, vol. 22, no. 1, pp. 191-202, 2014. 2. J. Yu and W. C. Wong, "Optimal Association in Wireless Mesh Networks" IEEE Trans. on Vehicular Technology, vol. 64, no. 5, pp. 2084-2096, 2015. 3. M. Derakhshani, X. Wang, T. Le-Ngoc, and A. Leon-Garcia, Airtime usage control in virtualized multi-cell 802.11 networks in IEEE Globecom Workshops, Dec 2015, pp. 1 6. 4. M. Al-Rodhaan, Q. Ma, and A. Al-Dhelaan, Genetic algorithm based approach to enhance network performance in multi-rate WLANs" Trans. on Networks and Comm., vol. 3, no. 6, p. 44, 2016. 5/13
Problem Formulation Dynamic Association Control in Virtualized Multi-rate WLANs Mied Integer Non-Linear Problem (NP-hard) Building the Rate Matri Solving the Relaed Optimization 2-Approimation Algorithm & Rounding 6/13
Meta-heuristic Solution Dynamic Association Control in Virtualized Multi-rate WLANs ACO-PF (Ant Colony Optimization for Proportional Fairness) Positive feedback-driven mechanism that evolves over time for converging to the best solution. using Meta-heuristic Solution to find a near-optimal Solution for maimizing PF Attractiveness (Heuristic) minimize Cost Total allocated BW to all the stations 1 / Attractiveness TotalBW = TotalBw + log (bi) 7/13
15 Users Meta-heuristic Solution Dynamic Association Control in Virtualized Multi-rate WLANs 6 APs 1 1 2 2 2 2 2 3 3 3 3 4 4 4 5 ACO-PF (Ant Colony Optimization for Proportional Fairness) Positive feedback-driven mechanism that evolves over time for converging to the best solution. using Meta-heuristic Solution to find a near-optimal Solution for maimizing PF Attractiveness (Heuristic) minimize Cost TotalBW = TotalBw + log (bi) Total allocated BW to all the stations 1 / Attractiveness 8/13
Evaluation Scenarios Eperiment Setup Protocol Interference Model ITU model for indoor attenuation Hotspot vs. Uniform user distribution Downlink greedy vs. Uplink greedy users 2.4 GHz (802.11g) Orthogonal APs Co-Channel APs 9/13
Selected Results Hotspot/Uplink/Co-channel Hotspot/Downlink/Co-channel 10/13
Test-bed Implementation Ali 3D2 - Compe 11/13
Selected Results Test-bed Results for UDP/TCP Traffic UDP Throughput ISP Load Sharing for RSSI & ACO-PF RSSI TCP Throughput ACO-PF 12/13
Summary Presenting an Adaptive & Cooperative Association Control Scheme for Common WLAN Infrastructures with Fairness Provisioning among Wi-Fi Stations Contributions Formulation of Wi-Fi Cooperative Association over Multiple ISPs Development of a Metaheuristic Scheme to Solve the Problem Faster & Less-Compleity Test-bed Implementation/Evaluation of the Proposed Scheme Ongoing/Future Works Etension of the current test-bed to SDN-enabled network devices Investigating the performance of ACO-PF for the heterogeneous traffic Integration of the proposed scheme with multi-hop Wi-Fi mesh networks 13/13
Q & A sajjadi @ uvic.ca http://seeb.ca
CAPWAP (Control And Provisioning of Wireless Access Points) RFC 5415 not implemented in an Interoperable way It has to forward all the Application Traffic to the Controller and it makes a Bottleneck. SDN provides a Common Language to control all Vendors' Equipment with higher level of granularity for fine-tuned control of network resources. It is a miture of Data, Control and Management planes that are bundled into a controller
hotspot downlink orthogonal uniform downlink orthogonal
6 APs 15 Users TotalBW = TotalBw + log (bi) Proposed Solution (ACO_PF) Attractiveness (Heuristic) TotalBW Cost 1 / TotalBW minimize 1 1 2 2 2 2 2 3 3 3 3 4 4 4 5
ACO Recap Ant Colony Optimization (ACO) a meta-heuristic optimization technique Proposed by Marco Dorigo in 1992 Food Nest
Stations Stations Stations Non-Linear Approimation Optimization 2 4 3 6 5 8 rate = APs 0.0000 1.7976 2.3593 3.1457 0.0000 0.0000 0.0000 0.0000 0.0000 3.5951 0.0000 4.7186 0.0000 1.5729 0.0000 0.0000 0.0000 0.0000 2.3593 3.1457 2.3593 2.3593 0.0000 0.0000 0.0000 0.0000 0.0000 1.5729 2.3593 4.7186 0.0000 1.7976 0.0000 0.0000 2.3593 0.0000 4.7186 1.5729 1.7976 0.0000 0.0000 0.0000 0.0000 2.3593 1.5729 4.7186 0.0000 1.7976 3.5951 0.0000 6.2915 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 6.2915 0.0000 3.5951 0.0000 0.0000 0.0000 0.0000 4.7186 0.0000 3.5951 0.0000 2.6963 0.0000 3.1457 0.0000 2.3593 0.0000 0.0000 0.0000 1 7 t = APs 0.0000 0.0000 0.0000 0.4999 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 = 9.2350 9.2522 9.5099 9.7920 9.5908 9.8507 9.2462 9.3360 0.9060 0.9996 0.9981 1.0000 0.9049 0.9659 0.9076 0.9076 0.9020 1.0000 0.9019 0.9998 0.9008 0.9958 0.9034 0.9036 0.9060 0.9084 0.9981 1.0000 0.9971 0.9982 0.9076 0.9077 0.9061 0.9084 0.9059 0.9978 0.9971 1.0000 0.9076 0.9988 0.9045 0.9068 0.9964 0.9596 1.0000 0.9967 0.9985 0.9060 0.9061 0.9084 0.9059 0.9990 0.9946 1.0000 0.9076 0.9988 0.9995 0.9039 1.0000 0.9584 0.9005 0.9644 0.9031 0.9032 0.9028 0.9051 0.9027 0.9589 0.9017 0.9997 0.9043 1.0000 0.9020 0.9043 0.9020 0.9586 0.9980 0.9645 1.0000 0.9036 1.0000 0.9074 0.9989 0.9598 0.9960 0.9655 0.9066 0.9067 = 10
Non-Linear Approimation Optimization Stations Stations Stations BP_mat = Virtual Slots of AP j Stations Sum of the elements of each row is 1 0.9060 0.0940 0 0 0 0 0 0 0 0 0 0.8080 0.1920 0 0 0 0 0 0 0 0 0 0.7140 0.286 0 0 0 0 0 0 0 0 0 0.6201 0.3799 0 0 0 0 0 0 0 0 0 0.5246 0.4754 0 0 0 0 0 0 0 0 0 0.4306 0.5694 0 0 0 0 0 0 0 0 0 0.4301 0.5699 0 0 0 0 0 0 0 0 0 0.3329 0.6671 0 0 0 0 0 0 0 0 0 0.2350 0.765.235 0.9996 0.0004 0 0 0 0 0 0 0 0 0 0.9996 0.0004 0 0 0 0 0 0 0 0 0 0.9080 0.0920 0 0 0 0 0 0 0 0 0 0.8163 0.1837 0 0 0 0 0 0 0 0 0 0.7231 0.2769 0 0 0 0 0 0 0 0 0 0.6314 0.3686 0 0 0 0 0 0 0 0 0 0.5354 0.4646 0 0 0 0 0 0 0 0 0 0.4405 0.5595 0 0 0 0 0 0 0 0 0 0.3449 0.6551.2522 0.9981 0.0019 0 0 0 0 0 0 0 0 0 0.9000 0.1000 0 0 0 0 0 0 0 0 0 0.8981 0.1019 0 0 0 0 0 0 0 0 0 0.8040 0.1960 0 0 0 0 0 0 0 0 0 0.8004 0.1996 0 0 0 0 0 0 0 0 0 0.7063 0.2937 0 0 0 0 0 0 0 0 0 0.7063 0.2937 0 0 0 0 0 0 0 0 0 0.6090 0.3910 0 0 0 0 0 0 0 0 0 0.5110 0.4890.5099 1.000 0 0 0.9998 0.0002 0 0 0 0 0 0 0 0 0 0.9998 0.0002 0 0 0 0 0 0 0 0 0 0.9976 0.0024 0 0 0 0 0 0 0 0 0 0.9572 0.0428 0 0 0 0 0 0 0 0 0 0.9562 0.0438 0 0 0 0 0 0 0 0 0 0.9147 0.0853 0 0 0 0 0 0 0 0 0 0.8736 0.1264 0 0 0 0 0 0 0 0 0 0.8322 0.1678.7920 80 10... 10 10 10 10 rate = t = = APs 0.0000 1.7976 2.3593 3.1457 0.0000 0.0000 0.0000 0.0000 0.0000 3.5951 0.0000 4.7186 0.0000 1.5729 0.0000 0.0000 0.0000 0.0000 2.3593 3.1457 2.3593 2.3593 0.0000 0.0000 0.0000 0.0000 0.0000 1.5729 2.3593 4.7186 0.0000 1.7976 0.0000 0.0000 2.3593 0.0000 4.7186 1.5729 1.7976 0.0000 0.0000 0.0000 0.0000 2.3593 1.5729 4.7186 0.0000 1.7976 3.5951 0.0000 6.2915 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 6.2915 0.0000 3.5951 0.0000 0.0000 0.0000 0.0000 4.7186 0.0000 3.5951 0.0000 2.6963 0.0000 3.1457 0.0000 2.3593 0.0000 0.0000 0.0000 APs 0.0000 0.0000 0.0000 0.4999 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 9.2350 9.2522 9.5099 9.7920 9.5908 9.8507 9.2462 9.3360 0.9060 0.9996 0.9981 1.0000 0.9049 0.9659 0.9076 0.9076 0.9020 1.0000 0.9019 0.9998 0.9008 0.9958 0.9034 0.9036 0.9060 0.9084 0.9981 1.0000 0.9971 0.9982 0.9076 0.9077 0.9061 0.9084 0.9059 0.9978 0.9971 1.0000 0.9076 0.9988 0.9045 0.9068 0.9964 0.9596 1.0000 0.9967 0.9985 0.9060 0.9061 0.9084 0.9059 0.9990 0.9946 1.0000 0.9076 0.9988 0.9995 0.9039 1.0000 0.9584 0.9005 0.9644 0.9031 0.9032 0.9028 0.9051 0.9027 0.9589 0.9017 0.9997 0.9043 1.0000 0.9020 0.9043 0.9020 0.9586 0.9980 0.9645 1.0000 0.9036 1.0000 0.9074 0.9989 0.9598 0.9960 0.9655 0.9066 0.9067
Non-Linear Approimation Optimization BP_mat = Virtual Slots of AP j Stations Sum of the elements of each row is 1 0.9060 0.0940 0 0 0 0 0 0 0 0 0 0.8080 0.1920 0 0 0 0 0 0 0 0 0 0.7140 0.286 0 0 0 0 0 0 0 0 0 0.6201 0.3799 0 0 0 0 0 0 0 0 0 0.5246 0.4754 0 0 0 0 0 0 0 0 0 0.4306 0.5694 0 0 0 0 0 0 0 0 0 0.4301 0.5699 0 0 0 0 0 0 0 0 0 0.3329 0.6671 0 0 0 0 0 0 0 0 0 0.2350 0.765.235 0.9996 0.0004 0 0 0 0 0 0 0 0 0 0.9996 0.0004 0 0 0 0 0 0 0 0 0 0.9080 0.0920 0 0 0 0 0 0 0 0 0 0.8163 0.1837 0 0 0 0 0 0 0 0 0 0.7231 0.2769 0 0 0 0 0 0 0 0 0 0.6314 0.3686 0 0 0 0 0 0 0 0 0 0.5354 0.4646 0 0 0 0 0 0 0 0 0 0.4405 0.5595 0 0 0 0 0 0 0 0 0 0.3449 0.6551.2522 0.9981 0.0019 0 0 0 0 0 0 0 0 0 0.9000 0.1000 0 0 0 0 0 0 0 0 0 0.8981 0.1019 0 0 0 0 0 0 0 0 0 0.8040 0.1960 0 0 0 0 0 0 0 0 0 0.8004 0.1996 0 0 0 0 0 0 0 0 0 0.7063 0.2937 0 0 0 0 0 0 0 0 0 0.7063 0.2937 0 0 0 0 0 0 0 0 0 0.6090 0.3910 0 0 0 0 0 0 0 0 0 0.5110 0.4890.5099 1.000 0 0 0.9998 0.0002 0 0 0 0 0 0 0 0 0 0.9998 0.0002 0 0 0 0 0 0 0 0 0 0.9976 0.0024 0 0 0 0 0 0 0 0 0 0.9572 0.0428 0 0 0 0 0 0 0 0 0 0.9562 0.0438 0 0 0 0 0 0 0 0 0 0.9147 0.0853 0 0 0 0 0 0 0 0 0 0.8736 0.1264 0 0 0 0 0 0 0 0 0 0.8322 0.1678.7920... 10 10 10 10 Profit_mat = 80 10 of each column 80 10 Profit of each edge = 0 0 0 0 0 0-5.0660 0 0 0 0 0 0 0 0 0-5.0660 0 0 0 0 0 0 0 0 0 0 0 0 14.8074 0 0 0 0 0 0 0 0 0 14.8074-5.5794 15.0951 0 0 0 0 0 0 0 0 0 15.0951 0 0 0 0 0 0 0 0-6.8204 0 0 0 0 0 0 0 0 0 0 0-6.8205 0 0 0 0 0 0 0 0 0-6.8205 0 0 0 0 0 0 0 0 0 0 0-7.4866 0 0 0 0 0 0 0 0 0-7.4866 0 0 0 0 0 0 0 0 0 0 0 15.6547 0 0 0 0 0 0 0 0 0 15.6547 0 0 0 0 0 0 0 0 0 0 0 0-6.0241 0 0 0 0 0 0 0 0 0-6.0241 14.2681 0 0 0 0 0 0 0 0 0 14.2681-4.7276 14.2687 0 0 0 0 0 0 0 0 0 14.2687-8.2607 0 0 0 0 0 0 0 0 0-8.2607 0 0 0 0 0 0 0 0 0 0 0-7.0584 0 0 0 0 0 0 0 0 0-7.0584 0 0 0 0 Pick the Ma Element...
Non-Linear Approimation Optimization 2 1 4 3 U1 = 1.50 U2 = 3.42 U3 = 1.50 U4 = 2.25 U5 = 4.50 U6 = 2.25 U7 = 6.00 U8 = 3.42 U9 = 3.42 U10=2.57 6 5 Sorted users w.r.t. Allocated BW 7 U7 = 6.00 U5 = 4.50 U2 = 3.42 U8 = 3.42 U9 = 3.42 U10=2.57 U4 = 2.25 U6 = 2.25 U1 = 1.50 U3 = 1.50 8 Profit_mat = Profit of each edge = 0 0 0 0 0 0-5.0660 0 0 0 0 0 0 0 0 0-5.0660 0 0 0 0 0 0 0 0 0 0 0 0 14.8074 0 0 0 0 0 0 0 0 0 14.8074-5.5794 15.0951 0 0 0 0 0 0 0 0 0 15.0951 0 0 0 0 0 0 0 0-6.8204 0 0 0 0 0 0 0 0 0 0 0-6.8205 0 0 0 0 0 0 0 0 0-6.8205 0 0 0 0 0 0 0 0 0 0 0-7.4866 0 0 0 0 0 0 0 0 0-7.4866 0 0 0 0 0 0 0 0 0 0 0 15.6547 0 0 0 0 0 0 0 0 0 15.6547 0 0 0 0 0 0 0 0 0 0 0 0-6.0241 0 0 0 0 0 0 0 0 0-6.0241 14.2681 0 0 0 0 0 0 0 0 0 14.2681-4.7276 14.2687 0 0 0 0 0 0 0 0 0 14.2687-8.2607 0 0 0 0 0 0 0 0 0-8.2607 0 0 0 0 0 0 0 0 0 0 0-7.0584 0 0 0 0 0 0 0 0 0-7.0584 0 0 0 0 Pick the Ma Element of each column 80 10...